Algorithm for learning negotiation strategy with reinforcement learning

摘要

In this paper, we propose an algorithm for self-bidding agents in the internet auction to learn bidding strategy depending on the circumstances. Q-learning and ε-greedy methods, the typical techniques in reinforcement learning, are used in learning and decision-making modules which are the basis of this algorithm. We applied this algorithm to an ascending-bid auction, and the agents could acquire the bidding strategy to get higher utility in one successful bid.